Curved sensors will solve some of the challenges posed by the laws of physics that manufacturers battle with when designing lenses, but when is this new technology going to bring about dramatic changes to the world of photography?
In this short video, Dave McKeegan explains the practical, scientific reasons that curved sensors make more sense than conventional flat ones, and discusses why this groundbreaking shift might not happen in the very near future. In addition, if and when they come to market, the transition might be slow, especially when you consider that entirely new systems will need to be designed and manufacturers prefer to wait around 30 years before launching a brand new lens mount — especially one that will have no backwards compatibility.
The Petzval field curvature inherent to today’s lenses — i.e., the curved image plane — isn’t the only difficulty posted by flat sensors, by the way. Colors have different points of focus which lens manufacturers are constantly trying to correct.
While chatting to engineer and camera guru Dave Haynie, he suggested that one option to achieve interchangeable lenses is to bundle the lens and the sensor together, as seen with the Ricoh GXR released back in 2009. Given that the cost of sensors is constantly falling, this might not be such an outlandish idea.
Will curved sensors change our cameras? Let us know your thoughts in the comments below.
While the advantages are real, the improvements in IQ will scarcely be significant enough to deploy in the enthusiast field.
Most likely they would find the best application in technical and scientific fields.
Of course if it becomes the pet issue of the noisy folks in the internet we may yet see them. Then we will have a whole new spate of pixel peeping of the tiny differences and a howling about price.
I think the biggest benefit to the consumer market is not in image quality (although there will definitely be better IQ), but in cost. If a sensor can be made with a spherical surface, then lenses with simple spherical elements can be used. Field correction and chromatic aberration correction for a flat surface accounts for much of the complexity of present lenses. The errors would be reduced significantly with a curved sensor, and thus compensatory optic design is much simpler, which translates to manufacturing ease with less exotic materials and processes. I am guessing that it would also reduce the complexity of autofocus and in-lens image stabilization systems as well.
The question then becomes whether or not a curved sensor be made cheaply enough to take advantage of the reduced lens costs. Given that any commercially viable curved sensor will have to be dynamically adjustable to account for lens focal length (radius of the focus sphere is the focal length of the lens), it may still be a tall hill to climb. But with the continuing evolution of MEMS (micro electro-mechanical systems) technology, it may become viable in the not too distant future, Fifteen years ago, MEMS rate sensors and accelerometers for orientation establishment and image stabilization were incredibly expensive, but now every cell phone has them and they are inexpensive.
It would seem that a curved sensor would have to be matched to the lens as the radius of curvature for a given FL is unique. Once you deviate from that you have one sensor that is perfect for X lens and every other FL has to be corrected thus losing th eadvantage.
If the sensor is fabricated to a fixed curvature, you would be correct. For this to be commercially viable, in my opinion, the sensor must be deformable to be able to adjust its curvature based upon focal length. On a large scale, self-compensating optics do this routinely by using faceted surfaces that are individually actuated. At a sensor level, it can conceivably be done, as I mentioned, with MEMS devices built into or onto the sensor.
When you say "large scale" are you talking about telescope optics?
If so, sure. But IMO the chance of a deformable sensor for even MF is nearly nil.
I've worked on advanced R&D prototype systems in which the focal plane sensor is faceted to allow for aberration correction. These were not for cameras, per se, but for imaging systems that were not terribly larger than a camera. albeit not the resolution of a camera sensor..
I agree that deformable silicon using present technology, given the amount of curvature that would be needed for a FF sensor with a wide angle lens is a very tall ask, and a "straightforward" implementation is hard to envision. But piezo motors to deform silicon has been commonplace for a long time.
Alternately, there are hybrid approaches that are in principal within technical reach. The curvature of the focus surface doesn't necessarily imply a curvature of the sensor surface, if the focus surface is not the pixel, but a transport medium. To picture this, think on a large scale - imagine the sensing surface is an array of individual optical fibers , one for each pixel, each with a nominal length plus slack, and a little linear actuator that can translate it longitudinally. The incoming rays are focus onto the ends of the fiber-optics, which then transport them down to the sensor elements on a flat array. varying the position of the business end of the fiber adjusts the focus surface. This is just for illustration purposes-not that one would ever do this in a camera. This type of system has been available for years on other types of imaging systems.
For a camera sensor, it might be possible to use the same concept at micro and nano scales using deformable silicon nitride membranes, or even something similar to the Texas Instruments digital micro mirror devices (DMD's). The sensor chip could be fabricated with a nominal curvature and variation due to focal length could be accomplished with the transport medium. These types of devices are not expensive. The DMD's are used in those compact digital projectors that you can connect to your tablet or smartphone, which cost less than $100.
All that being said, the path from a working, producible prototype to a mass produced, reliable consumer product is longer than what most people would casually assume because every improvement has unintended consequences. For instance, it may make IBIS much trickier because the motion of the stabilization now has to account for a spherical field, instead of a flat field. But, maybe this would be compensated in the transport medium, or sensor deformation. My guess is there are a host of other little things like this that would need to be addressed. So, I really would be amazed if we saw a viable system in the next five years.
But then again, if this is the next major technology leap that the camera companies are looking to (I have no idea whether or not it is), they undoubtedly have researchers with much better ideas and knowledge of other cutting edge technologies that may be more viable.
All of this is just fun speculation.
While it may seem an interesting topic, It's risky to follow suggestions from a photography enthusiast youtuber without lens engineering background. E.g. if one reads further in the topic, you'd easily find articles and patents for variable curvature sensors for zoom lenses. It's not only about replacing a range of existing lenses, there's many more issues and limitations when designing such sensors and lenses.
Doesn’t the curvature of the sensor need to be matched with the lens? I think it’s only practical for fixed lens cameras. Even a zoom lens would have different field curvatures at different focal lengths.
Yes. A long lens projects a flatter image than a wide lens.
That's what I though as well.
So for a zoom lens, you'd even need the curvature to change as you zoom in or out. And when changing lenses from your 800mm zoom to your 15mm UWA, well whoops, whole different curvature of the sensor needed.
Building sensors into lenses as someone suggested still doesn't solve the zoomlens problem.
Even adjusting focus would change it (a bit).
All in all, that gives me the feeling that curved sensors also need to be flexible sensors and they will not be the next big breakthrough that some people seem to be expecting, not any time soon.
I like the sensor/ lens combo idea. What is the per chip cost for a modern full frame sensor? Like if Fuji wants to buy a 61mpx full frame sensor from Sony what is the price?
I thought that sensors were one of the expensive components? Hence FF and MF having such a price premium?
Besides, it still doesn't address the problem of zoom lenses needing different curvature at different zoom levels.
And ... how would you upgrade your sensor independently of your lens?
What would the interface between lens and camera look like at this point? The datarate would have to be extremely high.
I could only foresee this working for extremely high-end setups.
All the technical wizardry will never make up for lack of talent and work ethic.
Or, to paraphrase Ansel Adams "the world doesn't need more technically perfect boring pictures".